The Algorithm Study On Bearings-Only Target Localization And Tracking Using Two Or More Observers

Bearings-only target localization and tracking is an important branch of passive localization and tracking technique. In practical surveillance scenario, because the kinds of feature data measured from target are very limited, furthermore, when the target bearing becomes the only reliable information, the bearings-only target tracking technique is preferred to estimate target motion parameters, such as position, velocity, acceleration, etc. the traditional bearings-only target localization and tracking algorithm mainly focuses on the batch processing technique suffering from computational burden. Furthermore, when the pseudolinear estimator is adopted, the results are biased. Therefore, due to the tracking performance and the requirement of real time, this estimator is not satisfactory when applied to the case of tracking scenario. According to the imminent requirement of modern war, a practical, fast and reliable bearings-only tracking algorithm is preferred. Especially, the research on recursive-processing-technique-based bearings-only tracking algorithms has some practical meaning.Because the target localization (tracking) can be realised using multiple obsrevers at any given time index, moreover, there is no additional motion requirement for single observer, the multiple observer system has the advantage of strong ability of anti-jamming and survival. With the development of computer technique and modern communication technique, the multiple observer system (especially for two observer system) has been widely used in practical tracking scenario. In this paper, the bearings-only target tracking algorithm using two or more observers is studied, and the main contributions are list as follows:The observability problem of bearings-only target tracking using two observers is studied, and two cases are considered, i.e., there exists system error or not at all.If system error does not exist, a nearly unbiased parameter estimation method, which is based on the minimization of expectation of mean square equivalent error obtained from the pseudo-linearization of measurement equation, is proposed. Otherwise, two algorithms are proposed, i.e, a batch-processing-technique-based a new instrumental variable method and dual filter strategy, respectively.